The observation that democracies rarely if ever fight each other was made by Dean Babst nearly three decades ago, but has had little impact on the literature on peace research and international relations until recently. But now every volume of the leading journals contains articles on minor and major aspects of this theme. Professional jealousy and confusion of levels of analysis are possible explanations for the late acceptance of the idea of a democratic peace, but above all it seems to have been hampered by the Cold War. Erich Weede has taken a bold step in reconsidering his own previous view and other should follow. The Cold War has ended in the real world, and it should end in peace research, too.

‘Democracy encourages peaceful interaction among states’.Footnote 1 This proposition flourished during the Enlightenment—it was, for instance, a central part of the political debates which surrounded the American and French revolutions. As early as 1795 Immanuel Kant described a ‘pacific federation’ or ‘pacific union’ created by liberal republics.Footnote 2 Much more recently this topic has become the subject of systematic empirical observation. In this issue of JPR we publish four articles on the relationship between democracy and peace. In the first, Weede (1992) reconsiders his previously published view (1984, 1989) that extended deterrence and subordination to superpowers are the major pacifying conditions in the international system. He now joins the emerging consensus that ‘democracies do not fight each other’, that democracies have established a ‘separate peace’. Forsythe (1992), while accepting this conclusion, has a different main concern: to investigate a semi-deviant case, how democracies may substitute covert action for overt force against popularly elected governments which pursue policies strongly disliked by the United States or other major democratic powers. Sørensen (1992) accepts the Kantian vision, while wishing to retain some basic insights of neorealism. Russett/Antholis (1992) attempt to extend the coverage of the proposition that democracies rarely fight each other to the ancient Greek city-states. Two issues ago, Starr (1992) sought to link relatively recent empirical findings about the lack of war among democracies to more established theoretical ideas about pluralistic security communities (Deutsch et al. 1957).

These are but a few examples from a burgeoning literature. Every recent volume of the leading journals in international relations and peace research contains articles on some major or minor aspect of this theme.

The observation that democracies do not fight each other was made almost three decades ago by Babst (1964, 1972).Footnote 3 Babst had examined data on 116 major wars from 1789 to 1941 from Wright (1965) and found that ‘no wars have been fought between independent nations with elective governments’ (1964: 10). Applying a probabilistic argument to the two world wars of this century, he concluded that it was extremely unlikely all the elective governments (10 out of the 33 independent nations participating in World War I; 14 out of 52 in World War II) should be on the same side purely by chance.

A second round of debate was initiated by Rummel (1983), who argued that ‘libertarian’ states were more peaceful and that libertarian states never fought each other.Footnote 4 His argument quickly led to rejoinders by Chan (1984), Weede (1984), and others. At the same time, Doyle (1983, 1986) was developing an argument based on the views of Kant.

After nearly a decade of debate following Doyle’s and Rummel’s articles, there is now a near-consensus on two points: that there is little difference in the amount of war participation between democracies and non-democracies (Rummel being the major dissenter here) but that wars (or even military conflicts short of war) are non-existent (or very rare) among democracies. Indeed, several scholars have echoed Levy’s statement that this ‘absence of war between democratic states comes as close as anything we have to an empirical law in international relations’ (Levy 1989: 270). This empirical regularity has never been seriously called into question.

Enthusiasm for this remarkable finding should be tempered with an appreciation that it applies only in cases where a relatively high threshold is set for both ‘democracy’ and ‘war’. Take democracy first: Most scholars have followed (more or less) the criteria carefully specified by Small/Singer (1976): (a) free elections with opposition parties, (b) a minimum suffrage (10 %); and (c) a parliament either in control of the executive or at least enjoying parity with it. Schweller (1992: 240) applauds Doyle’s definition of ‘representative government’ with the suffrage level raised to at least 30 % (and female suffrage granted within a generation of its initial demand) and further requires the government to be (d) ‘internally sovereign over military and foreign affairs’ and (e) stable (in existence for at least 3 years). He also adds (f) individual civil rights and—more controversially perhaps—(g) private property and a free-enterprise economy. Attempts to lower the thresholds in empirical studies have not been successful: During the Peloponnesian Wars, ‘democracies were slightly more likely to fight one another than to fight any other type of regime’ (Russett/Antholis 1992: 424) even though a norm was found to be emerging among democracies against fighting other democracies. Using cross-cultural ethnographic evidence from the Human Relations Area Files, Ember et al. (1992) found more supportive evidence, but the hypothesis had to be substantially revised to be testable on these data.Footnote 5 Even in the modern age, lowering the suffrage threshold makes an anomalous case of the British-American war of 1812.Footnote 6

Lowering the threshold for war below the 1,000 battle deaths used in the Correlates of War datasets on international and extrasystemic wars also produces less clear-cut results. Maoz/Abdolali (1989) and Maoz/Russett (1992) have tested propositions about democracy and war on the dataset on ‘militarized interstate disputes’, also generated within the COW project (Gochman/Maoz 1984). In the latest of these studies, 15 cases of disputes between democracies have to be accounted for. To forestall criticism that war is so rare an event that ‘it is difficult to demonstrate the effectiveness of pacifying conditions’ (p. 380), Weede extends his study, too, to the militarized disputes dataset. This yields one such dispute between two democracies, Finland and Norway! Weede fails to find this dispute in other comparable datasets and raises questions about the coding scheme of the militarized disputes data. However, there was in fact a dispute between Finland and Norway in 1976–77 (about German NATO forces in Norway) and this discussion also referred to the friendship treaty between Finland and the USSR, which might be invoked in the case of a new threat from Germany. It may or may not be reasonable to characterize this as a ‘militarized dispute’; in any case, such incidents are so far from war that it is unreasonable to assume they should be accounted for by the same factors.Footnote 7

‘Democracies don’t fight each other’—why was such a simple observation not made in the great classical studies of war? Richardson (1960) did not touch this topic at all. Wright (1965) dealt at some length with the relationship between democracy and war, but did not comment on the lack of war between democracies.Footnote 8 And why, when this striking regularity was noticed in the early 1960s, did it take nearly thirty years before it became widely acknowledged?

One answer to the latter question may be professional jealousy. Babst was a criminologist and generally published in journals which must be regarded as extremely obscure from a peace research point of view. Nevertheless, his second article was spotted by professional students of war. In a frequently quoted article, Small/Singer (1976: 51) lampooned Babst’s finding: ‘In a less academic enterprise, a recent issue … prominently featured an analysis that allegedly ends the debate [on regime-type and foreign conflict behavior] forever.’ Such a ‘seductive proposition’ was likely ‘to be accepted uncritically by those searching for some ray of hope in the generally bleak picture of contemporary international relations’. Since they found some of Babst’s coding rules to be ‘invisible’ they generated a new dataset from their own Correlates of War project in order to examine Babst’s ‘superficially credible proposition’.

A second reason for the late acceptance may be methodological. A number of early contributions to the literature confuse the issue of a national-level proposition (democracies are more peaceful) and a dyadic-level proposition (democracies don’t fight each other). Babst argued strictly in terms of the latter (as Kant had done, 169 years earlier), but Singer & Small, in their polemic against Babst, set out to demolish in some detail a proposition of the first type, ‘the innate peacefulness of the bourgeois democracies’. Thus, Small & Singer really knocked down a straw man. But so persuasive was their article that for a long while no one pursued this lead. When Rummel joined the battle he chose to defend the very thesis that Small & Singer had disconfirmed, that ‘libertarian’ states were inherently more peaceful. This drew attention away from his second thesis on dyadic peace between libertarian states.

Despite the increasing methodological sophistication of research on international violence, the difference between a main effect and an interaction effect is not always grasped. Moreover, with an increasing number of nations, the idea of conducting research at the dyadic level—where the number of units of analysis is roughly the square of the number of nations—is not especially appealing, either to research directors or funding agencies.

While the concept of an interaction effect may be too complicated, the finding that democracies don’t fight each other may also be seen as too simple, even simplistic. In the midst of regression analyses, factor analyses, and numerous other multivariate techniques, the idea that one variable alone is a sufficient (but not necessary) condition for a state of peace in the sense of non-war seems ridiculously naive. For instance, towards the end of their article Small/Singer (1976: 67) did admit that they could only find two very marginal cases of ‘bourgeois democracies’ fighting each other, but they dismissed this ‘superficial proof of the innate peacefulness of the bourgeois democracies’ with a comment that this might perhaps be accounted for by geographical proximity, since wars tended to be fought between neighbors and few democracies had common borders. This is only the first of many attempts to explain away the idea of a separate peace among democracies with reference to third variables—many of which have recently been put to rest by Maoz/Russett (1992). Testing for third variables brings the issue into the normal grind of research practice, but such tests for ‘statistical artifacts’ are less meaningful in the case of perfect or near-perfect correlations.Footnote 9 Regardless of the variables used to subdivide a population of nation-pairs, a zero in the cell for joint democracy will remain zero in any subdivision. It is possible, of course, that a third variable might be found which would account for joint democracy as well as for nonwar. But the only third variables which could perform such a feat would themselves need to have a perfect relationship with both the other variables. A little reflection should suffice to show that geographical distance is not such a variable: most wars have been between neighbors, but certainly not all. And although most democracies are not neighbors, some are. In fact, in more than 25 years of research on ‘Correlates of War’ no one has come up with any relationship nearly as strong as the dyadic relationship between democracy and nonwar. Therefore, it seems extremely unlikely that such an underlying causal factor will be identified. Of course, it will not be hard to find separate third-variable explanations for each separate peace, deterrence here, distance there, and so on. Such explanations will be advanced with particular fervor by those hostile to any quantitative analysis.Footnote 10 In fact, by their very diversity they do little to bolster the many armchair generalizations, frequently single-factor ones, about the war-making of democracies.

As far as third variables are concerned, the perfect or near-perfect correlation between democracy and nonwar in dyads should soon begin to have a very different effect: all research on the causes of war in modern times will be regarded as suspect if it is not first corrected for this factor. In fact, I would argue that most behavioral research on conditions for war and peace in the modern world can now be thrown on the scrap-heap of history, and researchers can start all over again on a new basis. Despite mental resistance to such an idea, this is exactly as it should be in a cumulative discipline. A similar caution must be exercised in formulating new hypotheses about war. For instance, a number of authors are currently urging that environmental problems are a major factor in causing war.Footnote 11 This general thesis seems extremely implausible if it is meant to include war over environmental issues between democratic countries. As Diehl (1992: 340) points out, some relationships are so powerful that they supersede any other conditions for war. The proper approach here would be first to sort out the double democratic dyads and then to look at the environmental factors in outbreaks of war in the remaining dyads.Footnote 12 Ten years from now, the finding that democratic countries don’t war with each other will probably be regarded as extremely trivial in a research design—a factor to be corrected for before we get on with the real job of accounting for the wars that do occur. By then, ‘antipositivists’ who now reject the democracy-nonwar relationship may revert to their second line of defense, that quantitative research can do nothing but belabor the obvious.

A further reason for discounting the dyadic relationship between democracy and peace was that it seemed to be based, on the one hand, on rather raw empiricism (Babst, Rummel) and, on the other, on airy philosophical principles (Kant, Doyle). Schweller (1992: 235) argues that much of the literature is ‘data-driven’ and Lake (1992: 24) feels that ‘No theory presently exists that can account for this striking empirical regularity’. While this criticism might have been true in the 1970s, it would no longer seem to hold. Fairly elaborate theoretical arguments have been made in terms of constraints on decision-makers in democracies, in terms of democracy as an exercise in non-violent domestic conflict resolution which can be extended to international affairs if a suitable (i.e. democratic) counterpart is found, in terms of democracies seeing the mutual relationships as positive- sum rather than zero-sum, or in terms of state rent-seeking, which creates an imperialist bias in a country’s foreign policy, but less so in democracies.Footnote 13 While there is as yet no consensus on which theoretical rationale accounts best for the observed relationship—or on how to separate them empirically—there is at least no lack of convincing theories.

The apparent discrepancy between findings at the nation level and at the dyadic level also calls for explanation. Theoretically, of course, the two can easily be linked: the simplest way to do so is to assume that non-democratic nations tend to attack peaceful democratic nations and that the wars fought by democratic countries are always defensive. In this way, the war participation of the democracies becomes as high as that of the non-democracies, even though the former are more peaceful. Small/Singer (1976: 66), however, found no differences between democratic and non-democratic countries with respect to war initiation. More recently, Schweller (1992: 249) hypothesized ‘that only authoritarian regimes initiate preventive war and that they do so regardless of whether the challenger is democratic or authoritarian’, he found the empirical evidence to be ‘overwhelming’ from Sparta to Nazi Germany.Footnote 14 Declining democratic leaders tend to seek accommodation when faced by democratic challengers, and a defensive alliance when challenged by a nondemocracy. Schweller found Israel to be the leading candidate for a deviant case from the latter regularity—one, however, which does not contradict the main regularity that we are discussing here. Morgan/Schwebach (1992: 312) also concluded ‘that democratic states are less likely to escalate disputes than are non-democracies’. Lake (1992: 30), on the other hand, maintains that ‘democracies are not only less likely to wage war with each other’, but that ‘they are also significantly more likely to win the wars they fight against autocracies’, a regularity which no doubt has contributed to skepticism about the peacefulness of democracies. The debate on this and other theoretical issues will obviously continue—hopefully some of it will take place in this journal!

A fifth reason for not taking much account of the democracy-war relationship at the dyadic level is that when it was first proposed by Kant there were only three liberal regimes in existence (Switzerland, France, and the USA; Doyle 1986: 1164). Thus, Kant’s writings might be dismissed as theoretical speculation about a hypothetical future world with no empirical evidence and without much consequence in a world of despots. In the two centuries since then a ‘separate peace’ has spread to an increasing number of states: roughly 50 for the period since 1945, according to Doyle. Not only are there more democracies around, but their numbers are increasing. When 10 % of the world’s nations were democracies (roughly the state of affairs in the 19th century) only close to 1 % of all nation-pairs were excluded from war.Footnote 15 With 50 % democracies—not an unrealistic target for the close of this century—the separate peace encompasses close to 25 % of all pairs. This, then, is the basis for the ‘obsolescence of war in the developed world’ heralded by Mueller (1989).

And finally, a more political explanation for the tardy response of the research community to the idea of the separate democratic peace: Virtually all systematic research concerned with causes of war has taken place in countries affected by the Cold War. Research attributing major importance to political democracy seemed propagandistic to many peace researchers who subscribed to a ‘third way’ in the Cold War and disliked anything that smacked of one-sided propaganda for ‘the free world’.Footnote 16 Babst’s original article was not entirely free of this preaching when it suggested that democracy was a great force for peace and that ‘diplomatic efforts at war prevention might well be directed toward further accelerating’ the growth of elective governments. Small/Singer (1976: 51, n. 3) suggested, however, that Babst’s prescription ‘could, paradoxically enough, turn out to be a major stimulus to war’, an observation compatible with at least some of the rhetoric in the 1991 Gulf War. Among the potentially important policy implications of Rummel’s work on this topic, Vincent (1987: 104) singled out one he clearly regarded as unsavory: ‘that American covert and overt interventions for the purpose of democratizing a society would help promote peace in the world system’. The debate about imperialism in the 1970s focused, unsurprisingly, more on the war-mongering nature of several democracies than on their peacefulness. But the idea of a democratic separate peace seemed too soft for the realists, who felt more comfortable with deterrence and strict bipolarity (and still do, as is evidenced by the doomsday predictions for Europe after the Cold War in a celebrated article by Mearsheimer 1990).Footnote 17 As a former member of the deterrence school of thought, Weede has taken a bold step in reconsidering his own views. Peace researchers who rejected the link between democracy and peace from a radically different paradigm should not be less forthright. The strong finding about the ‘democratic peace’ may to some extent have been a victim of the Cold War. No wonder then that it fell to an ‘innocent criminologist’ to observe that the emperors had no clothes. The Cold War has now ended in the real world; it should end in peace research, too.